Description

After upgrading to bun v1.2.0, the column ordering in generated SQL tables has changed when using embedded structs. Previously, columns were created in the order they were defined in the struct (including embedded structs). However, in v1.2.0, the columns from embedded structs appear after the directly defined columns, regardless of their position in the struct definition.

Reproduction

// 20210505110026_foo.go 
package main
type BaseModel struct {
	ID        int
	CreatedAt time.Time `bun:",default:current_timestamp"`
}
type FooModel struct {
	BaseModel

	Name string
}

func init() {
	Migrations.MustRegister(func(ctx context.Context, db *bun.DB) error {
		_, err := db.NewCreateTable().Model((*FooModel)(nil)).Exec(ctx)
		return err
	}, func(ctx context.Context, db *bun.DB) error {
		_, err := db.NewDropTable().Model((*FooModel)(nil)).Exec(ctx)
		return err
	})
}

// main.go
package main
import (
	"context"
	"database/sql"

	"github.com/uptrace/bun"
	"github.com/uptrace/bun/dialect/sqlitedialect"
	"github.com/uptrace/bun/driver/sqliteshim"
	"github.com/uptrace/bun/migrate"
)

var Migrations = migrate.NewMigrations()

func main() {
	sqldb, err := sql.Open(sqliteshim.ShimName, "file:test.s3db?cache=shared")
	if err != nil {
		panic(err)
	}

	db := bun.NewDB(sqldb, sqlitedialect.New())
	if err := Migrations.DiscoverCaller(); err != nil {
		panic(err)
	}
	migrator := migrate.NewMigrator(db, Migrations)

	ctx := context.Background()
	err = migrator.Init(ctx)
	if err != nil {
		panic(err)
	}

	_, err = migrator.Migrate(ctx)
	if err != nil {
		panic(err)
	}
}

Expected Behavior

The generated table should have columns in the order they are defined in the struct hierarchy:

1. id
2. created_at
3. name

Actual Behavior

After v1.2.0, the generated table has columns in this order:

1. name
2. id
3. created_at

The columns from the embedded BaseModel struct are placed after the directly defined Name column, which breaks the expected structural order.

Version Information

• Before: Working as expected in versions prior to v1.2.0
• After: Column ordering changed in v1.2.0 and above

Impact

This change affects systems that rely on consistent column ordering, particularly in scenarios where:

• Database tools or ORMs assume a specific column order
• Migration scripts depend on predictable column positioning
• Legacy system integrations expect a particular column sequence

Additional Notes

The change in column ordering appears to be related to how bun v1.2.0 handles embedded struct fields during table creation. While this doesn't affect functionality, it represents a breaking change in terms of database schema generation.